Workers leaving declining occupations often land in another one, UCLA sociologists say
UCLA sociologist Jennie Brand's team reports in Science Advances that workers in declining occupations have worse upward mobility than those in growing ones. The AI-specific findings were published separately, in the American Economic Association's Papers and Proceedings.
The Scientist · Science desk
Declining-occupation job changes often land in another declining one. Exposed, declining jobs: job pressure; may face limited paths to expanding fields. Payroll, bookkeeping clerks: high exposure; under 1% list a generative AI skill. ML researchers: highly exposed, report AI skills.
- constraint Workers in declining occupations Changing occupations often moves them into another declining occupation without improving economic position, claim 7
- exposure Exposed, declining occupations Workers face pressure on existing jobs and may face limited mobility paths into expanding fields, claim 9
- exposure Payroll and bookkeeping clerks Preliminary research on AI skills identifies high AI exposure in these occupations, among others, claim 10
- constraint Payroll, bookkeeping clerks Less than 1% in the researchers' LinkedIn data list a generative AI skill on their profiles, claim 11
- capability Machine learning researchers Also highly exposed to AI; workers in these fields report more AI-related skills, claim 15
| Who | How | Kind | Claim |
|---|---|---|---|
| Workers in declining occupations | Changing occupations often moves them into another declining occupation without improving economic position | constraint | 7 |
| Exposed, declining occupations | Workers face pressure on existing jobs and may face limited mobility paths into expanding fields | exposure | 9 |
| Payroll and bookkeeping clerks | Preliminary research on AI skills identifies high AI exposure in these occupations, among others | exposure | 10 |
| Payroll, bookkeeping clerks | Less than 1% in the researchers' LinkedIn data list a generative AI skill on their profiles | constraint | 11 |
| Machine learning researchers | Also highly exposed to AI; workers in these fields report more AI-related skills | capability | 15 |
What happened
- Changing occupations often moves a worker into another declining occupation without improving economic position, which the authors call a mobility trap.
- Its preliminary work flags high exposure in payroll clerks, insurance claims clerks, bookkeepers, medical transcriptionists, data-entry workers, and interpreters and translators.
- In the team's LinkedIn data, fewer than 1% of payroll and bookkeeping clerks list a generative AI skill on their profiles.
Why it matters
- decision Anyone ranking roles for retraining by AI-exposure score alone is using a measure the authors say cannot identify who is most vulnerable.
- constraint The idea that AI deepens the trap remains a stated concern. The authors say their current research does not directly test every mechanism that would make exposed workers more vulnerable.
- exposure One exposure label covers payroll clerks with almost no listed AI skill and machine learning researchers who report more, so it groups very different workers together.
AI enters this research through a second axis, and the definition matters. The authors call an occupation AI-exposed when large language models and related software could cut the time for particular tasks by at least half while maintaining quality [4]. That measures what software could do to task time. Headcount is a separate question: an exposed occupation might shrink, grow or change substantially as workers use AI [5].
The trap and the exposure meet in one group. Brand and her co-author Xi Song of Columbia single out workers in occupations that are both highly exposed and projected to decline. These workers face pressure on their existing jobs and may have limited paths into expanding fields [9]. On the exposure score alone, Brand and Song said: "Exposure alone is insufficient to identify who is most vulnerable." [8] They also separate exposure from skill, since a job that AI can affect does not guarantee the skills or support to work effectively with it [6].
The skills evidence is thinner than the exposure evidence. It comes from what the authors call preliminary research [10], and from LinkedIn profiles, where fewer than 1% of payroll and bookkeeping clerks list a generative AI skill [11]. Machine learning researchers are also highly exposed but report more AI-related skills [15]. A profile line is a self-reported signal, and a clerk who uses these tools daily but has not listed them would not show up in that figure.
We would want to know whether workers in declining occupations differ in other ways from workers in growing ones. The Q&A does not report the sample, the size of the mobility gap or an effect size, so the Science Advances result reaches us as a direction without a magnitude.
The authors are explicit that movement alone proves little. A worker who switches occupations may still end up in a shrinking field, or in a job that pays less [13]. They also say people short on time or money, or without employer training or other jobs to move into, may be more vulnerable, though their current research does not directly test all these mechanisms [12].
What to watch
- Whether the American Economic Association Papers and Proceedings results report mobility outcomes for AI-exposed workers, with sample sizes and effect sizes.
- Whether the team directly tests the proposed mechanisms: time, money, employer training and access to alternative jobs.
- Whether the preliminary AI-skills work moves beyond LinkedIn profile listings as its measure of skill.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence50
- Adoption
- Insufficient
- Hype gap−10
- Incentives
- Insufficient
- Confidence50
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Jennie Brand, a UCLA professor of sociology, and her research partners at the Occupation Lab have been examining how occupational change, including AI exposure, affects workers' job mobility.
- [2]
In the Science Advances paper 'Trapped in declining occupations: Barriers to worker mobility in a changing economy', Brand's team found that workers in growing occupations have better prospects for upward mobility than those in declining occupations, creating a 'mobility trap' for the latter group.
- [3]
Findings specific to AI-exposed jobs were also published in the Papers and Proceedings of the American Economic Association.
- [4]
The researchers consider a job AI-exposed if large language models and related software could reduce the time needed for particular tasks by at least half while maintaining quality.
- [5]
The number of people in an occupation highly exposed to AI might decline, grow or change substantially as workers use AI.
- [6]
The researchers distinguish AI exposure from AI skills: having a job that AI can affect does not necessarily mean having the skills or support to work effectively with it.
- [7]
Changing occupations often results in moving into another declining occupation without improving economic position, which the authors call a 'mobility trap'.
- [8]
"Exposure alone is insufficient to identify who is most vulnerable."
- [9]
The authors highlight workers in occupations that are both highly exposed to AI and projected to decline in number; these workers face pressure on existing jobs and may face limited mobility paths into expanding fields.
- [10]
Preliminary research on AI skills identifies high AI exposure in occupations such as payroll clerks, insurance claims and policy-processing clerks, bookkeeping clerks, medical transcriptionists, data-entry workers, and interpreters and translators.
- [11]
Less than 1% of payroll and bookkeeping clerks in the researchers' LinkedIn data list a generative AI skill on their profiles.
- [12]
Vulnerability may be greater for people who lack time, money, employer training or access to alternative jobs, although the current research does not directly test all these mechanisms.
- [13]
Movement itself does not show whether people are getting ahead: someone can change occupations yet remain in a declining field or move into lower-paid work.
- [14]
Xi Song, who received her Ph.D. from UCLA, is now a professor at Columbia University and answered the Q&A questions alongside Brand.
- [15]
Some scientific and technical occupations, such as machine learning researchers, are also highly exposed to AI, but workers in these fields report more AI-related skills.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- phys.orgJob loss and AI-exposed jobs: Q&A with sociology professor
1 article · October 14, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- AI Labor MarketFollow
- Occupational mobilityFollow
- AI exposureFollow
Entities
- Jennie BrandFollow
- Xi SongFollow
- Occupation LabFollow
- Science AdvancesFollow
- University of California, Los AngelesFollow